Config Management Camp 2024 Ghent

Fixing Salt's atomic file access problem
2024-02-06, 16:45–17:10, B.2.015

Explore how Salt's approach to file management has room for improvement with learnings from a real-world at-scale production environment. You'll learn how to solve these problems "the SRE way", and see how well they work, until they don't. We'll discuss alternative solutions as provided by Salt as well as alternative solutions that can really scale.

In this presentation we'll deep-dive into various mechanisms that Salt provides out of the box to manage file access and versioning, like the gitfs fileserver, and how well each of them can work for various use-cases. Based on practical learnings, see how the "90%" SRE solution is good enough, until it isn't. Follow our journey toward a final solution that addresses all of the shortcomings of all of the previous solutions and how it's implemented.

I work at Cloudflare as an SRE, working closely with a large Salt deployment spanning a global production fleet of servers.